as of today i am a very long OpenWrt user and currently i am using a Netgear R6260 (with deactivated wifi) as my main router with sqm, DoH, HE ipv6 tunnel etc.... I am also have Netgear Orbi RBK750 configured as APs.
Currently i am using a VDSL line but soon i get a fiber-cable to my home with 400/80 Mbit. with the potential up to 1Gbit.
Now my question is: should i get a new router or will be the r6260 be enough.
If yes should i get a raspi4 with usb3-network dongle or should i better get an pcengines apu2e4 for future proof.
In the future i also want to implement a wireguard vpn on the router.
Short pro/con for me so far:
rpi4 + dongle pros: cheaper, could be used as other thing, more common
pcengins pros: x86 with aes-ni, 3 intel nics, nice looking case, buy and use
rpi cons: tinkering with case/powersupply, usb3 dongle, ...
pcengins cons: don't know anything about them, pricier, support?
How are the compareable performance-wise? How stable are the pcengins apu2
Do they sqm with potential 1Gbps?
Or are there any other options?
Thanks
limited to ~two interfaces and require smart switch for anything more extensible
limited to ~near gigabit
non mainline patches
for most SOHO / residential purposes more than enough... and for the price... expecting beyond 1.5-2years service life before being repurposed is unrealistic... ( given the rate of broadband technology this pretty much applies to most NTD / edge combos ) with exception of x64 or modular enterprise gear...
imho... it's really the third point... that is troublesome and probably should be a key metric in your decision...
It is possible to future proof, but the cost is high. My last router had a lifespan of 6 years, and was only replaced because the motherboard died, but it was costly to build.
Anything involving a "future-proof" spec today that would cause it to last you more than 2 years would need to be an x86_64 with multiple cores and 4x 10Gbe interfaces with a healthy chunk of RAM. Probably not worth the price unless you plan to make use of a good amount of its power today.
I am using an apu2e4 as firewall running OPNsense (not pfSense) and OpenWRT APs (GliNet B-1300). This gives me 800Mbit/s using wired (over NAT) and 866 Mb/s in 5ghz WIFI-AC.
In my other house, I am using only GLinet B-1300 hardware (OpenWRT) and I can reach 1Gb/s on wire.
Modern hardware like the GLiNet B1300 is enough to reach good speed. It is running on ARMv6 and consumes low energy. Most of the time, you don't need 1Gb/s.
Yes, I agree and I thought that I was making exactly that point.
Select modern hardware is good enough for 1Gbps, but to future proof for 4 years in the future will require more substantial hardware as speeds will very likely well exceed 1Gbps by then.
And I also made the point that it's not worth the money to future proof this way unless you plan to use a lot of that power today...
Thanks for the input. I meant „future proof“ for the potential upgrade to 1Gbps. For my village i don‘t think they would upgrade the fibre above 1Gbps in the next years, they currently rolling out the fiber and this is planned to be complete 2023.
I believe even an RPi4 could handle over 1Gbps with appropriate USB 2.5Gbps NIC and a LAG group in a decent small business switch. You LAG together the onboard NIC and the 2.5Gbps USB NIC, and then put VLAN tags, which gets you ~ 3.5 Gbps potential, which should be enough for 5 years even in areas with 1Gbps already.
The RPi should route 3Gbps in all likelihood (with appropriate packet steering across multiple CPU) and probably even SQM up to 2Gbps.
The truth of the matter though, is I'd take 500Mbps reliable symmetric with SQM over 1Gbps + with crap buffer management.
Unless you're heavy into SQM or VPN usage, the r6260[0] should cope with up to 1 GBit/s line speed (simple firewall rules, hardware flow-offloading enabled, no SQM) - so I'd just wait and see what happens once you're upgraded to ftth. Even if your current router wouldn't be able to cope, that doesn't imply you'd be without internet access - you 'just' won't reach full throughput available from your ISP and still have the time to go shopping then.
Given that throughputs above 300-500 MBit/s WAN speed still require rather high-end devices at rather inflated prices (and OpenWrt support might not be there either), so a cheap solution that fully supports the intermediate needs right now is a better strategy than attempting to future proof the router for hopeful speed increases in the future. The RPi4 (or e.g. the NanoPi r4s) could be rather budget friendly solutions for now (which won't be a big financial loss if you need to upgrade/ replace them in the future), x86_64 with 10GBASE-T or SFP+ 10 GBit/s ethernet cards the most future proof alternative (while that is rather affordable, 10 GBit/s capable switches are not - yet).
--
[0] I certainly wouldn't recommend to buy a mt7621a device with that in mind, but if you already have it at your disposal 'for free' (as in already paid for), it's certainly worth testing first, before shelling out more money.
+1; I think that is the best approach, only upgrade once you see how your current router performs; that should give you a better idea if you need "slightly faster" or massively faster.
Side note, I concur with @dlakelan, that depending on your use-case SQM at way below link and contractual rate can still result in a more useful internet access. Case in point, I had been traffic shaping my 100/40 VDSL link down to 49/32 as that was the highest I could get SQM to shape in that router, after testing that 100/40 without SQM behaves less well than 49/32 with SQM (my measure here is how well interactive uses behave with heavy up- and/or downloads happening at the same time). Your mileage will vary, and that really depends on your use-cases, so slower with SQM is not necessarily better than faster without, but it might be worth a try.
